[Purpose/significance] Accurately grasping the driving factors and irrational performance characteristics of different types of social media users' differential situational motivational states under the risk perception of opinion in⁃volvement, and improving the targeting of the regulation of irrational user behavior and online public opinion risk man⁃agement. [Method/process] Based on the semi-structured interview data, user labels are extracted by using the root⁃ing theory, and a social media user portrait model of irrational behavior under the risk perception of public opinion in⁃volvement is constructed under the framework of the situation theory of problem solving, and the group structure of four categories of user portraits with significant differences is derived through the K-means clustering algorithm for empiri⁃cal analysis. [Result/conclusion] The irrational behavior user groups under the perception of risk of public opinion in⁃volvement can be divided into four categories: information-addicted, passive-adherent, interest-focused, and lossavoidance,and according to the characteristics of the different dimensional labels, behavioral guidance strategies can be proposed from the four aspects of demand guidance, information regulation, emotional reshaping, and technological empowerment.